Video Recommendation System Dynamic Weight Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video recommendation systems face delays in model updating, leading to insufficient or excessive distribution of user interests due to inadequate control over real-time user preferences.
Innovation Solution
A method that acquires video tags from user history data, determines tag weights, calculates video weights for candidate videos, selects target videos based on these weights, and dynamically adjusts weights of undistributed videos to prevent dominance by a single point of interest, ensuring timely and effective distribution of user interests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the recommendation system uses traditional model updating methods, then the system structure remains simple, but the model updating is delayed and cannot capture real-time user preferences
Solution Approach 1:
The patent segments the recommendation system into multiple independent modules: candidate video generation module, tag weight determination module, video weight calculation module, and dynamic weight adjustment module. Each module operates independently with specific functions, allowing real-time processing without requiring complete model retraining, thus achieving fast updates while maintaining manageable system complexity
Solution Approach 2:
The system performs preliminary actions by pre-calculating and storing video tags and their weightings in advance. When user interactions occur, the system only needs to retrieve and adjust these pre-computed values rather than recalculating from scratch, enabling real-time responsiveness without complex real-time computation
2Ease of operation
If the recommendation system distributes videos based on single dominant interest, then the distribution process is simple, but the user experience deteriorates due to screen dominance by a single point of interest
Solution Approach 1:
The patent applies local quality by assigning different weight values to different video tags based on their relevance to user interests. Instead of uniform distribution or single-dominance approach, each video tag receives localized weight adjustment based on its specific contribution to user preference, achieving balanced and diverse video distribution that improves user experience
Solution Approach 2:
The system dynamically changes the weight parameters of video tags based on real-time user feedback and interaction data. By adjusting these parameters continuously, the system can adapt to changing user preferences and prevent any single point of interest from dominating the recommendation screen, maintaining both operational simplicity and high reliability
3Measurement precision
If the system calculates video weights based on comprehensive tag analysis, then the recommendation precision is improved, but the computational time increases
Solution Approach 1:
The system performs preliminary extraction and storage of video tags during video ingestion, so that when recommendation is needed, only weight calculation and comparison are required rather than full tag analysis. This pre-processing step significantly reduces computational time while maintaining precise video weight calculation through subsequent tag-based scoring
Solution Approach 2:
The patent uses tag representations as simplified copies of complex video content characteristics. Instead of analyzing entire video content in real-time, the system works with compact tag representations that capture essential features, achieving precise video weight calculation with minimal computational overhead
Data Source
AI summary
Provided is a method for recommending videos. The method includes: acquiring an identifier of a video operated by a target account in history, and determining a tag weight of each of a plurality of video tags of the video corresponding to the identifier; determining, based on the tag weight of the each video tag, a video weight of each of a plurality of candidate videos in a predetermined candidate set; selecting, based on the video weight of the each candidate video, a target video from the candidate set for recommendation; and detecting, from video tags of a distributed video, a video tag that meets a predetermined condition, and reducing, based on the tag weight of the video tag that meets the predetermined condition, a weight of an undistributed candidate video with the video tag that meets the predetermined condition in the candidate set.